CN111324698A

CN111324698A - Deep learning method, evaluation viewpoint extraction method, device and system

Info

Publication number: CN111324698A
Application number: CN202010104388.5A
Authority: CN
Inventors: 林坡; 沈艺; 陈述; 许加书; 梁诗雯
Original assignee: Suning Cloud Computing Co Ltd
Current assignee: Suning Cloud Computing Co Ltd
Priority date: 2020-02-20
Filing date: 2020-02-20
Publication date: 2020-06-23
Anticipated expiration: 2040-02-20
Also published as: CN111324698B

Abstract

The invention discloses a deep learning method, an evaluation viewpoint extraction method, a device and a system, wherein the extraction method comprises the following steps: performing word segmentation processing on evaluation information to be extracted to obtain word segmentation results; judging whether a target attribute word and a target emotion word which are respectively matched with at least one attribute word and at least one emotion word in a preset dictionary table exist in the word segmentation result; if yes, extracting target attribute words and target emotion words and generating feature data; the characteristic data comprises word distances between the target attribute words and the target emotion words; inputting the characteristic data into a deep learning model to obtain a modification relation between the target attribute words and the target emotion words; determining a target category of evaluation information to be extracted and determining a target polarity of a target emotion word; an evaluation viewpoint is formed according to the target type, the target polarity and the modification relationship. The method can accurately and efficiently extract the viewpoint of evaluation information.

Description

Deep learning method, evaluation viewpoint extraction method, device and system

Technical Field

The application relates to the technical field of information, in particular to a deep learning method, an evaluation viewpoint extraction device and an evaluation viewpoint extraction system.

Background

In online shopping, consumers generally judge that online merchandise values are not worth purchasing through evaluation information of other consumers.

However, negative contents such as rough quality and the like are often found in the text part of the evaluation information at present, but the content and the grade are not matched due to the condition that the grade evaluation is good. In order to correctly evaluate the rank of the evaluation information in terms of the character portion, it is necessary to analyze the evaluation information to extract the evaluation viewpoint.

In processing evaluation information, sentences are accurately segmented according to data characteristics of the evaluation information through a precision mode (jieba) of ending segmentation words, and words in natural language are converted into dense vectors which can be understood by a computer through word vectors. Defining a tag set of sequence labels using a BIO notation, AP being an abbreviation for Aspect, wherein B denotes the beginning of a rating viewpoint attribute word, and B _ AP being an abbreviation for Begin of Aspect; i represents the middle of an evaluation viewpoint attribute word, and I _ AP is an abbreviation of Inside of Aspect; o denotes Other non-appraisal opinion attribute words, i.e. the abbreviation of Other, different blocks can be assigned different labels by the three labels B _ AP, I _ AP and O.

A deep learning model based on a memory network is adopted, the memory network is merged into a two-way long and short memory network (Me-BilSTM), and comment viewpoint modes existing in historical sentences are extracted and stored and applied to unknown comment sentences.

Firstly, extracting the front and back dependency information between words in a sentence by using the BilSTM, then compressing the obtained front and back dependency information into a sentence characterization vector, on one hand, extracting the stored related comment sentence mode vector from the memory network by using the characterization vector, and on the other hand, storing the currently processed sentence mode into the memory network. And finally, fusing the front and back dependency information of the sentence with the matched sentence mode vector, and performing final sentence labeling work by using a conditional random field.

① excessively depends on data purity, the problem of data imbalance is not considered, the requirement on the data volume of attribute words and emotion words in the training Chinese evaluation data is too high, sufficient Chinese data volume is generally required to be obtained, otherwise, the accuracy rate is low, ② training speed is too low, the length of Chinese comment sentences is too long, the accuracy rate is low, ③ cannot analyze newly appeared attribute words and emotion words.

Therefore, a more efficient and accurate method for analyzing the evaluation information is needed to extract the evaluation viewpoint.

Disclosure of Invention

The invention aims to provide a deep learning method, an evaluation viewpoint extraction method and an evaluation viewpoint extraction system aiming at the defects of the prior art, wherein evaluation information is analyzed and judged by a training model, and then the evaluation viewpoint is extracted efficiently and accurately.

The invention discloses a deep learning method, which comprises the following steps:

performing word segmentation processing, part of speech tagging and word distance tagging on the sample evaluation information to obtain a sample word segmentation result;

extracting sample attribute words and sample emotion words in the sample word segmentation result, and generating sample characteristic data by combining word distances between the sample attribute words and the sample emotion words;

labeling sample modification relations between sample attribute words and sample emotion words in the sample word segmentation results;

and training to obtain a deep learning model by taking the sample characteristic data as input and the sample modification relation as output.

Preferably, the extracting sample attribute words and sample emotion words in the sample word segmentation result includes:

and inquiring in a word bank table containing corresponding relations between words and parts of speech, and extracting the sample attribute words and the sample emotion words in the sample word segmentation result.

The invention also discloses an evaluation viewpoint extraction method, which comprises the following steps:

performing word segmentation processing and part-of-speech tagging on evaluation information to be extracted to obtain word segmentation results;

extracting target attribute words and target emotion words in the word segmentation result, combining the target attribute words and the target emotion words, and generating feature data according to word distances between the target attribute words and the target emotion words;

inputting the feature data into the deep learning model to obtain a modification relation between the target attribute words and the target emotion words;

determining the target category of the evaluation information to be extracted and determining the target polarity of the target emotional words;

and forming the evaluation viewpoint according to the target class, the target polarity and the modification relation.

Preferably, the method further comprises:

and if the word segmentation result contains the target emotion words and no target attribute words, matching the corresponding target attribute words for the target emotion words according to the preset correspondence between the emotion words and the attribute words.

Preferably, the determining the target polarity of the target emotion word comprises:

if the target emotion words and the target attribute words are matched with a target fixed collocation polarity in a fixed collocation polarity corresponding relationship, determining the polarity of the target emotion words according to the target fixed collocation polarity; the fixed collocation polarity corresponding relationship comprises a corresponding relationship of fixed collocation of at least one pair of polarity, emotion words and attribute words;

and if the target emotion words and the target attribute words are not matched with the target fixed collocation polarity in the fixed collocation polarity corresponding relation, determining the target polarity of the target emotion words according to the preset emotion word and polarity corresponding relation.

Preferably, the determining the target category of the evaluation information to be extracted includes:

and determining at least one target category corresponding to the target attribute words according to the preset corresponding relation between the attribute words and the evaluation information categories.

Preferably, the target category is at least one of goods, logistics, services and price.

The invention also discloses a deep learning device, comprising:

the sample word segmentation unit is used for carrying out word segmentation processing, part of speech tagging and word distance tagging on the sample evaluation information to obtain a sample word segmentation result;

the sample feature data unit is used for extracting sample attribute words and sample emotion words in the sample word segmentation result and generating sample feature data by combining word distances between the sample attribute words and the sample emotion words;

the sample modification relation labeling unit is used for labeling the sample modification relation between the sample attribute words and the sample emotion words in the sample word segmentation result;

and the model training unit is used for training to obtain a deep learning model by taking the sample characteristic data as input and the sample modification relation as output.

The invention also discloses an evaluation viewpoint extraction device, which applies the deep learning model and comprises:

the word segmentation unit is used for performing word segmentation processing and part-of-speech tagging on the evaluation information to be extracted to obtain a word segmentation result;

the feature data unit is used for extracting target attribute words and target emotion words in the word segmentation result, combining the target attribute words and the target emotion words and generating feature data according to word distances between the target attribute words and the target emotion words;

the model prediction unit is used for inputting the feature data into the deep learning model to obtain a modification relation between the target attribute words and the target emotion words;

a category determination unit configured to determine a target category of the evaluation information to be extracted;

the polarity determining unit is used for determining the target polarity of the target emotional words;

an evaluation point of view unit for forming the evaluation point of view according to the target class, the target polarity, and the modification relationship.

Finally, a computer system is also disclosed, comprising:

one or more processors; and

a memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform the method described above.

The invention has the beneficial effects that:

the method can judge the modification relation between the attribute words and the emotion words in the evaluation information through the deep learning model, and further generates an evaluation viewpoint by combining polarity and category judgment. The model of the invention can judge only by carrying out word segmentation, part of speech tagging and word distance tagging on the evaluation information, has simple realization, high efficiency, no dependence on context and high accuracy, and can analyze new emotional words and attribute words, thereby further improving the accuracy of the evaluation viewpoint. The evaluation information can be well corresponded to the corresponding rating evaluation based on the extracted evaluation viewpoint, and the problem caused by wrong matching rating and evaluation content is avoided.

Drawings

The accompanying drawings, which are included to provide a further understanding of the embodiments of the application, are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. It is obvious that the drawings in the following description are only some embodiments of the application, and that for a person skilled in the art, other drawings can be derived from them without inventive effort. In the drawings:

FIGS. 1 and 2 are schematic diagrams of evaluation information labeling results of samples according to the present application;

FIG. 3 is a flowchart of the evaluation point extraction method of example 2 of the present application;

fig. 4 is a structural view of an evaluation point extraction device in example 4 of the present application;

FIG. 5 is a block diagram of a computer system according to the present application.

Detailed Description

In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The rating point of view generally needs to describe what emotional description is made for what attributes under what categories (product, logistics, service, price, etc.), whether this description is positive or negative.

For example, the user evaluation viewpoint extracted after the analysis is as follows: the quality of the product is coarse and negative. Wherein, the product is classified, the quality is attribute, the roughness is emotion description, and the polarity of the emotion is negative polarity.

For this purpose, it is necessary to analyze what kind of classification the above-mentioned elements in the evaluation information, i.e., what kind of attribute the evaluation information relates to, what kind of emotional description is made, and this description is positive or negative. The attribute and the emotion description relate to parts of speech, namely emotion words and attribute words, so word segmentation processing, part of speech tagging and extraction of emotion words and attribute words are required. What kind of attribute is described as what kind of emotion, and also relates to the identification of the modification relation between the emotion words and the attribute words, i.e. which emotion word is used to modify which attribute word.

The word segmentation and the like can be processed by means of the current common word segmentation tools such as ansj, and the emotion words and the attribute words can be extracted and matched through a preset word stock table. If a word bank table is preset for common words and parts of speech (emotional words and attribute words) in the evaluation field, the part of speech of each word after word segmentation is determined by matching the result of word segmentation of the evaluation information with the word bank table, so as to extract the emotional words and the attribute words.

Regarding the recognition of the modification relation between the emotion words and the attribute words, the application provides a deep learning method, and a model is trained by means of word distances between the emotion words and the attribute words for the subsequent recognition of the modification relation.

First, a sample set is prepared, the sample set being a certain number of pieces of sample evaluation information.

And then, performing word segmentation and labeling on the sample evaluation information, wherein the labeling comprises part-of-speech labeling, word distance labeling and modification relation labeling. Sample attribute words and sample emotion words in the word library table can be extracted by means of the preset word library table containing the corresponding relation of common words and parts of speech (emotion words and attribute words) in the field. The word distance can be determined by preset index values of each participle, and the specific implementation can be realized by means of a related method of the word distance in the prior art. The modification relationship is classified into a modified relationship and an unmodified relationship, and can be specifically expressed by RIGHT (modified relationship) and relationship _ WRONG (modified relationship). As shown in fig. 1 and 2, the processed data features include attribute words a, emotion words S, non-attribute words O, word distance, and modification relation descriptions.

And inputting the labeling information into a preset model, taking part-of-speech labels and word distance labels as input, taking modification relation labels as output, performing deep learning, and training to obtain the model.

The model can be specifically a Textcnn model based on a deep learning classification model, and the Textcnn model can judge whether the emotion words and the attribute words in the evaluation information have modification relations or not through training.

The algorithm pre-research part is divided into a training part and a verification part, and the proportion of the training part to the verification part is 5: the neural network model part adopts convolution layer, activation layer, pooling layer and full connection layer, and uses 3 different convolution kernels.

Through the trained model, the modification relation between the emotion words and the attribute words in the evaluation information can be determined, and then an evaluation viewpoint is formed by combining other information such as the polarity of the emotion words, the classification of the attribute words and the like. The specific steps for forming an evaluation viewpoint by using the trained model are as follows:

firstly, loading the trained model and a configuration file used for operating the model, and performing word segmentation on evaluation information to be extracted to obtain a word segmentation result;

and then extracting attribute words and emotional words in the word segmentation result by using a preset word bank table containing corresponding relations of common words and parts of speech (emotional words and attribute words) in the field. Due to the word stock table, attribute words and emotion words can be added or modified in the past word stock table to be dynamically modified subsequently.

And if the attribute words and the emotion words are extracted, generating characteristic data according to word distances between the attribute words and the emotion words, inputting the characteristic data into the model, and outputting the modification relation between the attribute words and the emotion words by using the model.

And then determining the polarity of the emotional words and the category of the evaluation information, and generating an evaluation viewpoint by combining the modification relation.

The polarity of the emotion words can be determined by the following method:

if the target emotion words and the target attribute words are matched with a target fixed collocation polarity in a fixed collocation polarity corresponding relationship, determining the target polarity of the target emotion words according to the target fixed collocation polarity; the fixed collocation polarity corresponding relationship comprises a corresponding relationship of fixed collocation of at least one pair of polarity, emotion words and attribute words;

and if the target emotion words and the target attribute words are not matched with the target fixed collocation polarity in the fixed collocation polarity corresponding relation, determining the polarity of the emotion words according to the preset emotion word and polarity corresponding relation.

For example, words such as "service dissatisfaction", "poor quality", etc. appear in the evaluation information, and the correspondence between the preset emotion words and the polarities can be used to judge that the emotion words "dissatisfaction", "poor" are negative polarities; for example, words such as "service satisfaction" and "quality is found in the evaluation information, and the emotional words" satisfaction "and" good "are determined to be positive.

Consider that some words have a particular polarity in a particular scenario, such as simply in many cases the polarity is positive, but negative if the collocation package is simple. In addition, in the case of negative words, the polarity is reversed, and negative polarity is not necessarily intended. Therefore, the polarity under the condition can be determined according to the fixed collocation polarity corresponding relation.

For example, the packaging is simple, and the polarity is determined as the negative polarity by fixing the corresponding relationship of the collocation polarities. The service is not satisfied, the negative word, the polarity is reversed to negative.

The determination of the category of the evaluation information to be extracted may be performed by:

and determining at least one target category corresponding to the attribute words according to the preset corresponding relation between the attribute words and the evaluation information categories. The category may specifically be at least one of a commodity, a logistics, a service, and a price.

Such as: and if the attribute word is good in attitude, the attribute word can be matched with multi-class service and logistics.

It should be noted that, after matching with the word bank table, if there is an emotion word and there is no attribute word in the word segmentation result, the corresponding attribute word may be matched for the emotion word from the preset emotion and attribute relation word bank. If only the emotional word of 'cheap' appears in the evaluation information, adding the common attribute word 'price' corresponding to 'cheap', and finally waiting until the complete extraction information, such as price: cheap: 1.0.

Through the model training mode in the method, a model is trained, the modification relation between the attribute words and the emotion words can be judged according to the word distance between the extracted attribute words and the extracted emotion words, and then the evaluation viewpoint can be obtained by combining the category and the polarity. The method is simple to operate, does not need to depend on context, is high in accuracy, and can analyze and judge newly added emotional words and attribute words.

Example 1

To sum up, embodiment 1 of the present invention provides a deep learning method, including:

Preferably, the deep learning model is a TextCNN deep learning model.

Example 2

Corresponding to the model, embodiment 2 of the present invention further discloses an evaluation viewpoint extraction method, to which the model is applied, as shown in fig. 3, the method includes:

s31, performing word segmentation processing and part-of-speech tagging on the evaluation information to be extracted to obtain word segmentation results.

This step can be performed by means of a word segmentation tool such as ansj.

S32, extracting the target attribute words and the target emotion words in the word segmentation results, combining the target attribute words and the target emotion words, and generating feature data according to word distances between the target attribute words and the target emotion words.

And extracting the target attribute words and the target emotion words according to a preset word bank table. Such as common attribute words and emotion words in the field of statistical sorting evaluation, and sorting into a table. And performing matching filtering on the word segmentation result by using the table, and extracting target attribute words and target emotion words in the evaluation information.

Based on the form of the table, the attribute words and the emotion words in the table can be added, deleted and modified subsequently according to the needs.

The word distance is needed for the prediction of the model, and for this reason, the word distance between the target attribute word and the target emotion word can be further determined. The determination may be specifically made using an index value table. The word distance determination is a well-known solution in the prior art, and the present invention is not limited to this.

And generating characteristic data based on the extracted target attribute words, target emotion words and word distances, wherein the characteristic data serves as input parameters of the model.

S33, inputting the feature data into the deep learning model to obtain the modification relation between the target attribute words and the target emotion words.

A plurality of target attribute words and target emotion words may be extracted from the evaluation information, and modification relations between each target attribute word and each target emotion word can be judged one by one based on the model.

If the evaluation information is 'delivery is fast, packaging is firm, and quality is a bit rough', the judgment result is as follows: the fast and the shipment have a decorative relationship, the firm and the package have a decorative relationship, and the roughness and the quality have a decorative relationship.

S34, determining the target category of the evaluation information to be extracted and determining the target polarity of the target emotion words.

The category may include logistics, services, goods, price, etc., and the corresponding category may be determined by the correspondence of the attribute words to the category. If the attribute word is delivery, the category is logistics. The correspondence between the attribute words and the categories may be preset.

The polarity is the positive polarity and the negative polarity mentioned above, and can be determined by the emotional words.

S35 forming the evaluation viewpoint based on the object type, the object polarity, and the modification relationship.

The evaluation view may be in terms of "original data category (there are multiple categories): attribute words: emotional words: and an information intercepting part: the data structure of polarity "is expressed as: good quality, positive polarity. Of course, the polarity can be expressed numerically, e.g., a positive polarity of 1.0 and a negative polarity of-1.0, the above example can be expressed as a good product: good quality: 1.0.

In a preferred embodiment, the method further comprises:

because the evaluation information is input by the user and may not be standard, when matching is performed according to the word bank table, only the target emotion words but no target attribute words may be matched in the word segmentation result, and at this time, the corresponding target attribute words may be matched for the target emotion words according to the preset correspondence between the emotion words and the attribute words.

In a preferred embodiment, the determining the target polarity of the target emotion word includes:

In a preferred embodiment, the determining the target category of the evaluation information to be extracted includes:

Example 3

Corresponding to embodiment 1, the present invention also discloses a deep learning apparatus, including:

Example 4

Corresponding to the above embodiment 2, the present invention further discloses an evaluation viewpoint extracting apparatus, to which the above deep learning model is applied, as shown in fig. 4, the apparatus includes:

a word segmentation unit 41, configured to perform word segmentation processing and part-of-speech tagging on evaluation information to be extracted to obtain a word segmentation result;

a feature data unit 42, configured to extract a target attribute word and a target emotion word in the word segmentation result, combine the target attribute word and the target emotion word, and generate feature data according to a word distance between the target attribute word and the target emotion word;

a model prediction unit 43, configured to input the feature data into the deep learning model to obtain a modification relationship between the target attribute word and the target emotion word;

a category determination unit 44 configured to determine a target category of the evaluation information to be extracted;

a polarity determining unit 45, configured to determine a target polarity of the target emotion word;

an evaluation point of view unit 46 for forming the evaluation point of view based on the target class, the target polarity, and the modification relationship.

Example 5

Corresponding to embodiment 2, embodiment 5 of the present invention further discloses a computer system, including:

one or more processors; and

Fig. 5 illustrates an architecture of a computer system, which may include, in particular, a processor 1510, a video display adapter 1511, a disk drive 1512, an input/output interface 1513, a network interface 1514, and a memory 1520. The processor 1510, video display adapter 1511, disk drive 1512, input/output interface 1513, network interface 1514, and memory 1520 may be communicatively coupled via a communication bus 1530.

The processor 1510 may be implemented by a general-purpose CPU (Central processing Unit), a microprocessor, an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits, and is configured to execute related programs to implement the technical solution provided by the present invention.

The Memory 1520 may be implemented in the form of a ROM (Read Only Memory), a RAM (random access Memory), a static storage device, a dynamic storage device, or the like. The memory 1520 may store an operating system 1521 for controlling the operation of the computer system 1500, a Basic Input Output System (BIOS) for controlling low-level operations of the computer system 1500. In addition, a web browser 1523, a data storage management system 1524, an icon font processing system 1525, and the like can also be stored. The icon font processing system 1525 may be an application program that implements the operations of the foregoing steps in this embodiment of the present invention. In summary, when the technical solution provided by the present invention is implemented by software or firmware, the relevant program codes are stored in the memory 1520 and called for execution by the processor 1510.

The input/output interface 1513 is used for connecting an input/output module to realize information input and output. The i/o module may be configured as a component in a device (not shown) or may be external to the device to provide a corresponding function. The input devices may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output devices may include a display, a speaker, a vibrator, an indicator light, etc.

The network interface 1514 is used to connect a communication module (not shown) to enable the device to communicatively interact with other devices. The communication module can realize communication in a wired mode (such as USB, network cable and the like) and also can realize communication in a wireless mode (such as mobile network, WIFI, Bluetooth and the like).

The bus 1530 includes a path to transfer information between the various components of the device, such as the processor 1510, the video display adapter 1511, the disk drive 1512, the input/output interface 1513, the network interface 1514, and the memory 1520.

In addition, the computer system 1500 may also obtain information of specific extraction conditions from the virtual resource object extraction condition information database 1541 for performing condition judgment, and the like.

It should be noted that although the above devices only show the processor 1510, the video display adapter 1511, the disk drive 1512, the input/output interface 1513, the network interface 1514, the memory 1520, the bus 1530, etc., in a specific implementation, the devices may also include other components necessary for proper operation. Furthermore, it will be understood by those skilled in the art that the apparatus described above may also include only the components necessary to implement the inventive arrangements, and need not include all of the components shown in the figures.

From the above description of the embodiments, it is clear to those skilled in the art that the present invention can be implemented by software plus necessary general hardware platform. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which may be stored in a storage medium, such as a ROM/RAM, a magnetic disk, an optical disk, or the like, and includes several instructions for enabling a computer device (which may be a personal computer, a cloud server, or a network device) to execute the method according to the embodiments or some parts of the embodiments of the present invention.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the system or system embodiments are substantially similar to the method embodiments and therefore are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described system and system embodiments are only illustrative, wherein the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

The method, the device and the system provided by the invention are described in detail, the principle and the implementation mode of the invention are explained by applying specific examples, and the description of the examples is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.

Claims

1. A deep learning method, comprising the steps of:

2. The deep learning method of claim 1, wherein extracting sample attribute words and sample emotion words in the sample word segmentation result comprises:

3. An evaluation point extraction method, characterized by comprising the steps of:

4. An evaluation point extraction method according to claim 3, further comprising:

5. The opinion extraction method according to claim 3 or 4, wherein the determining the target polarity of the target emotion words comprises:

6. An evaluation viewpoint extracting method according to claim 5, wherein the determining of the target category of the evaluation information to be extracted includes:

7. An evaluation point extraction method according to claim 3,

the target category is at least one of goods, logistics, services, and price.

8. A deep learning apparatus, comprising:

9. An evaluation point extraction apparatus to which the deep learning model according to claim 1 is applied, characterized by comprising:

the feature data unit is used for extracting target attribute words and target emotion words in the word segmentation result and generating feature data by combining word distances between the target attribute words and the target emotion words;

10. A computer system, comprising:

one or more processors; and

a memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform the method of any of claims 3-7.